整了个脚本，希望以后cpu高的时候，能够一键定位~~~~

郑重声明：参数要根据自身业务调整，线上执行jstack 要谨慎，本文只提供demo, 不作为线上操作标准。由于操作失误产生的后果由使用方自己完全负责哈~~~

1 用法：
（1）scmp 登录机器
（2）创建getCpuUsageOfJavaThread.sh 文件。

  1. vi getCpuUsageOfJavaThread.sh

（3） 将如下代码copy进 getCpuUsageOfJavaThread.sh

  1. #!/bin/bash
  2. echo "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~welcom~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"
  3. echo "usage: sh getCpuUsageOfJavaThread.sh sleepTime execCount cpuUsageThreshold"
  4. echo "eg: sh getCpuUsageOfJavaThread.sh 20 5 1   -- every 20s execute jstack once, and all execute 5 times . if cpu usage bigger than 1% while print"
  5. echo "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~gogogo~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"
  6. sleepTime=$1
  7. execCount=$2
  8. cpuUsageThreshold=$3
  10. if [ ! -n "$1" ] ;then
  11.     echo "param sleepTime is null"
  12.     echo "eg: sh getCpuUsageOfJavaThread.sh 20 5 1   -- every 20s execute jstack, and execute 5 times all. if cpu usage bigger than 1% while print"
  13.     exit
  14. fi
  16. if [ ! -n "$2" ] ;then
  17.     echo "param execCount is null"
  18.     echo "eg: sh getCpuUsageOfJavaThread.sh 20 5 1   -- every 20s execute jstack, and execute 5 times all. if cpu usage bigger than 1% while print"
  19.     exit
  20. fi
  22. if [ ! -n "$3" ] ;then
  23.     echo "param cpuUsageThreshold is null"
  24.     echo "eg: sh getCpuUsageOfJavaThread.sh 20 5 1   -- every 20s execute jstack, and execute 5 times all. if cpu usage bigger than 1% while print"
  25.     exit
  26. fi
  28. if [ ! -d "cpuData" ]; then
  29.   mkdir cpuData
  30. fi
  33. cd cpuData
  35. n=0
  36. while (($n<$execCount))
  37. do
  38.     #获取cpu使用大于sleepTime% 的 java 线程：  
  39.     now=$(date "+%Y%m%d-%H%M%S")
  40.     mkdir ${now}
  41.     cd ${now}
  42.     javapid=$(jps | grep "jar" | awk '{print $1}')&&ps H -e -o pid,tid,pcpu,args --sort=-pcpu |  awk -v jpid="$javapid" -v cpuUsageThreshold="$cpuUsageThreshold" '{if ($1 == jpid && $3 > cpuUsageThreshold) print $0}' > ${now}_threadCpuGT${cpuUsageThreshold}
  44.     awk '{print $2,$3}' ${now}_threadCpuGT${cpuUsageThreshold} > ${now}_threadId
  46.     jstack ${javapid} > ${now}_fullstackInfo
  48.     while read tid_cpu;
  49.     do
  50.         # 将线程转换为十六进制
  51.         tid=`echo $tid_cpu | awk '{print $1}'`
  52.         cpu=`echo $tid_cpu | awk '{print $2}'`
  53.         threadId16=$(printf "0x%x\n" ${tid})
  54.         echo "================= 线程id为 threadId-${tid}， 对应十六进制为 ${threadId16} , cpu使用率为 ${cpu} %. 详细堆栈信息如下 ：=================" >> ${now}_final_result
  55.         #grep -A 30 "nid="${threadId16}" " ${now}_fullstackInfo >> ${now}_final_result
  56.         sed -n -e "/nid=${threadId16} /, /^$/ p" ${now}_fullstackInfo >> ${now}_final_result
  58.     done < ${now}_threadId
  59.     rm ${now}_threadCpuGT${cpuUsageThreshold}
  60.     rm ${now}_threadId
  62.     #执行次数加一
  63.     n=$((n+1))
  64.     echo "=========execute ${n} times========="
  65.     # sleep 几秒
  66.     sleep $sleepTime
  67.     # 跳出一层，方便执行下一次
  68.     cd ..
  69. done
  70. echo "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ have a good time,  bye ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"

（4）执行如下指令：

郑重声明：参数要根据自身业务调整，线上执行jstack 要谨慎，本文只提供demo, 不作为线上操作标准。由于操作失误产生的后果由使用方自己完全负责哈~~~

demo :

  1. sh getCpuUsageOfJavaThread.sh 20 5 1

表示： 每20秒执行一次 ps 跟 jstack , 总共执行5次，会选择cpu 使用大于1% 的线程进行进行输出。

（5）结果文件说明：
文件夹路径： cd cpuData/xxxxxx-yy （相对于getCpuUsageOfJavaThread.sh 文件所在的目录）

  1. 20201204-062028_final_result 为 展示线程，以及对应的cpu使用率 的最终结果文件。
  2. 20200303-172806_fullstackInfo    jstack 完全堆栈信息

图片示例：
image.png

结果实例：
image.png